Predicting College Football Games

2024 Season Predictions

Author

Phil Henrickson

Published

September 19, 2024

This page displays predictions for games in the 4 college football season season. These predictions come from a model built on historical college football play by play and game data in order to simulate upcoming games. All data is from collegefootballdata.com.

About the Model

These predictions come from a game prediction model trained on historical college football data. My game prediction model uses team-level estimates of offensive, defensive, and special teams efficiency from play by play data. I simulate each game thousands of times in order to estimate the home team’s win probability and expected margin of victory (or defeat).

Notably, I do not use betting lines or the predictions of other models as features in my game prediction model. This allows me to directly compare my performance against that of other models and betting lines.

Week 4 Predictions

The following table displays estimates and predictions for the upcoming week’s college football games.

About the Table

Quality indicates the quality of the teams involved in the game. This is the (harmonic) mean of the two team’s power ratings and is scaled to range from 0 to 100, with 100 indicating a game between highly rated opponents.

Interest indicates how interesting a game is expected to be based on the quality of the teams and how competitive the game is expected to be. I assign every game a competitive rating using a quadratic formula applied to Pr(HomeWin), which I use along with Quality in a weighted average to produce the Interest score. This ranges from 0 to 100, where 100 indicates a very competitive game between two highly rated opponents.

Pr(HomeWin) is the percentage of times that the home team won the game across every simulation.

Prediction is the average margin of victory for the winning team across all simulations.

Betting Lines

How do these predictions compare to betting lines? I compare my predicted spread against the opening spread from ESPN Bet.

The following table displays my prediction for each game compared to opening betting lines from various providers.

Results as of Week 4

How has the model performed to date?

The model predicts the probability that the home team will win and their expected margin of victory (or defeat). I assess the model’s performance by assessing the accuracy of its predictions (win/loss), the logloss of its probabilities, and the mean absolute error of the predicted margin.

The following plot shows my model’s predicted margin of victory vs the actual margin of victory for all games so far this season.

Season Model Results
season season_type games accuracy mn_log_loss mae
2024 regular 241 0.797 0.371 15.398

Model Results by Week
season season_type week games accuracy mn_log_loss mae
2024 regular 1 100 0.870 0.266 15.215
2024 regular 2 78 0.692 0.501 15.981
2024 regular 3 63 0.810 0.378 14.968

Previous Predictions

The following table displays the model’s predictions for previous weeks along with the actual result.

Against the Spread

How has the model performed against the spread? I compare my model’s predictions against betting lines from various providers.

season week .metric Bovada DraftKings ESPN Bet
2024 1 accuracy 0.558 0.535 0.506
2024 2 accuracy 0.521 0.529 0.449
2024 3 accuracy 0.462 0.538 0.460

The following table displays my model’s predictions to date against betting lines.

Upcoming Games

The following table displays predictions for all upcoming games based on the results of the season so far.